Dissimilation in Hispano-Romance Diminutive Suffixation

A highly productive derivational process, diminutive suffixation in Spanish (e.g., gatito ~ gatiko/gatico ‘little/well-known/beloved/awful cat’ < gato ‘cat’) has received much attention in the morphology–phonology interface literature. The present study contributes a novel comparative analysis of a dissimilatory alternation between diminutive suffix allomorphs -ito/a and -ico/a (-iko/a) across three Hispano-Romance varieties. In Judeo-Spanish, the voiceless dorsal stop [k] of default -iko/a dissimilates to coronal [t] after any dorsal segment [k, ɡ, ɡʷ, x, w] in the base-final syllable. In Colombian Spanish, the voiceless coronal stop [t] of default -ito/a dissimilates to dorsal [k] after only an identical [t] in the base-final syllable. By contrast, Castilian Spanish -ito/a does not dissimilate, thereby providing a baseline for comparison. All three varieties allow for optional iteration of the suffix, which conveys greater smallness or endearment than the simple diminutive, e.g., Castilian Spanish gatitito ‘little/beloved kitty’, without dissimilation. Iterated diminutives in Colombian Spanish show two patterns of dissimilation, which have not been fully acknowledged in the previous literature. For example, either (i) [it] and [ik] alternate to avoid adjacent identical syllable onsets, e.g., gat[ikitíko], or (ii) [it] is iterated until alternating with word-final [ik], e.g., gat[ititíko]. In all three Hispano-Romance varieties, base-final unstressed vowels are deleted before a vowel-initial diminutive suffix, followed by unstressed -o/a, and stress (indicated by an acute accent) is shifted rightward onto the penultimate syllable of the diminutive word. Vowel deletion and stress shift apply recursively in iterated diminutives. We propose an Optimality Theory analysis of these alternations in terms of suffix allomorphy that is phonologically conditioned by consonantal place dissimilation. The analysis is formalized as an interaction among constraints that enforce prosodic unmarkedness, output–output correspondence, allomorph preference, and similarity avoidance. We consider theoretical alternatives and compare our analysis to other recent proposals.